Replicating Session Initiation Protocol (SIP) dialog state across multiple servers in a server cluster can be a very expensive proposition, both in terms of processing power and licensing costs. Off the shelf converged containers generally support a clustering offer that has much higher licensing costs than the non-clustered version. Even these expensive clustering solutions often have a failing in being able to handle failovers in the midst of a SIP transaction. This is unfortunate, given that entire self-service interactions and queuing often take place using “early media” in the midst of a SIP transaction. This is especially true in Europe, where regulations exist that demand customers not pay long distance fees until they are talking to a live person.
One illustrative session reconstruction architecture is taught in U.S. Patent Publication No. 2014/0047122, the entire contents of which are hereby incorporated herein by reference. While the '122 publication addresses many of the shortcomings described above, it presents a few shortcomings of its own: (i) it requires a proxy that can alternate route requests and responses based on a fully qualified domain name; (ii) the SIP flow required between the front-end processor and container is somewhat complex; and (iii) the receipt of an incoming SIP message is the only stimulus that can trigger a reconstruction. In the case that the call is queued, however, the stimulus most likely to arrive is one from the agent selection application indicating that an agent is available. An alternate means of session reconstruction is required for this scenario.